Narrow network insurance plans are a popular choice among consumers buying health insurance in the ACA exchanges. Plans with narrow networks selectively contract healthcare providers, restrict provider choice and do not or partly cover care outside the network. Selective contracting is intended to contain costs in the health care system and to stimulate competition among insurance companies. If these plans can offer lower premiums with a network that provides meaningful access to care, they may offer value to consumers.

We do not know if and how much insurance companies actually save by restricting provider choice. The first objective of this dissertation was to understand whether narrow network insurance plans have reduced healthcare expenditures and to identify the mechanism by which that reduction occurred. Next, this dissertation modeled individual preferences regarding attributes of the network of primary care providers and estimated the willingness to pay for provider choice. Finally, consumers' sentiments towards health insurance, healthcare providers and choice were analyzed.

To model the effect of narrow network plans on expenditures, two-part models were used on expenditures and negative binomial models for visit and drugs counts. Enrollees of narrow network plans had $761 lower annual total expenditures than those in plans that cover care outside the network and a substantial amount of that reduction is due to price reductions through selective contracting practices, not by limiting access to care by fewer visits.

A discrete choice experiment was used to elicit consumers' preferences regarding attribute(s) of the provider network. Mixed logit models were used to account for individual preference heterogeneity and to estimate willingness to pay for these attributes. For consumers, coverage of a personal doctor, was the most important attribute of provider networks. On average, respondents were willing to pay $95 more to get an insurance plan that covers care for their personal doctor.

Finally, text mining and machine learning techniques were used to analyze consumers' attitudes towards health insurance plans, health providers and choice and to identify factors that are associated with switching. Consumers trusted their personal doctors and other medical providers but feared unanticipated health events and severe medical events.